IB/qib: Add logic for affinity hint

Call irq_set_affinity_hint() to give userspace programs such as
irqbalance the information to be able to distribute qib interrupts
appropriately.

The logic allocates all non-receive interrupts to the first CPU local
to the HCA.  Receive interrupts are allocated round robin starting
with the second CPU local to the HCA with potential wrap back to the
second CPU.

This patch also adds a refinement to the name registered for MSI-X
interrupts so that user level scripts can determine the device
associated with the IRQs when there are multiple HCAs with a
potentially different set of local CPUs.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

drivers/infiniband/hw/qib/qib.h

@@ -427,6 +427,14 @@ struct qib_verbs_txreq {
 /* how often we check for packet activity for "power on hours (in seconds) */
 #define ACTIVITY_TIMER 5
 
+#define MAX_NAME_SIZE 64
+struct qib_msix_entry {
+	struct msix_entry msix;
+	void *arg;
+	char name[MAX_NAME_SIZE];
+	cpumask_var_t mask;
+};
+
 /* Below is an opaque struct. Each chip (device) can maintain
  * private data needed for its operation, but not germane to the
  * rest of the driver.  For convenience, we define another that
@@ -1355,7 +1363,7 @@ int qib_pcie_init(struct pci_dev *, const struct pci_device_id *);
 int qib_pcie_ddinit(struct qib_devdata *, struct pci_dev *,
 		    const struct pci_device_id *);
 void qib_pcie_ddcleanup(struct qib_devdata *);
-int qib_pcie_params(struct qib_devdata *, u32, u32 *, struct msix_entry *);
+int qib_pcie_params(struct qib_devdata *, u32, u32 *, struct qib_msix_entry *);
 int qib_reinit_intr(struct qib_devdata *);
 void qib_enable_intx(struct pci_dev *);
 void qib_nomsi(struct qib_devdata *);

drivers/infiniband/hw/qib/qib_iba7322.c

@@ -541,8 +541,7 @@ struct qib_chip_specific {
 	u32 lastbuf_for_pio;
 	u32 stay_in_freeze;
 	u32 recovery_ports_initted;
-	struct msix_entry *msix_entries;
-	void  **msix_arg;
+	struct qib_msix_entry *msix_entries;
 	unsigned long *sendchkenable;
 	unsigned long *sendgrhchk;
 	unsigned long *sendibchk;
@@ -639,24 +638,24 @@ static struct {
 	int lsb;
 	int port; /* 0 if not port-specific, else port # */
 } irq_table[] = {
-	{ QIB_DRV_NAME, qib_7322intr, -1, 0 },
-	{ QIB_DRV_NAME " (buf avail)", qib_7322bufavail,
+	{ "", qib_7322intr, -1, 0 },
+	{ " (buf avail)", qib_7322bufavail,
 		SYM_LSB(IntStatus, SendBufAvail), 0 },
-	{ QIB_DRV_NAME " (sdma 0)", sdma_intr,
+	{ " (sdma 0)", sdma_intr,
 		SYM_LSB(IntStatus, SDmaInt_0), 1 },
-	{ QIB_DRV_NAME " (sdma 1)", sdma_intr,
+	{ " (sdma 1)", sdma_intr,
 		SYM_LSB(IntStatus, SDmaInt_1), 2 },
-	{ QIB_DRV_NAME " (sdmaI 0)", sdma_idle_intr,
+	{ " (sdmaI 0)", sdma_idle_intr,
 		SYM_LSB(IntStatus, SDmaIdleInt_0), 1 },
-	{ QIB_DRV_NAME " (sdmaI 1)", sdma_idle_intr,
+	{ " (sdmaI 1)", sdma_idle_intr,
 		SYM_LSB(IntStatus, SDmaIdleInt_1), 2 },
-	{ QIB_DRV_NAME " (sdmaP 0)", sdma_progress_intr,
+	{ " (sdmaP 0)", sdma_progress_intr,
 		SYM_LSB(IntStatus, SDmaProgressInt_0), 1 },
-	{ QIB_DRV_NAME " (sdmaP 1)", sdma_progress_intr,
+	{ " (sdmaP 1)", sdma_progress_intr,
 		SYM_LSB(IntStatus, SDmaProgressInt_1), 2 },
-	{ QIB_DRV_NAME " (sdmaC 0)", sdma_cleanup_intr,
+	{ " (sdmaC 0)", sdma_cleanup_intr,
 		SYM_LSB(IntStatus, SDmaCleanupDone_0), 1 },
-	{ QIB_DRV_NAME " (sdmaC 1)", sdma_cleanup_intr,
+	{ " (sdmaC 1)", sdma_cleanup_intr,
 		SYM_LSB(IntStatus, SDmaCleanupDone_1), 2 },
 };
 
@@ -2567,9 +2566,13 @@ static void qib_7322_nomsix(struct qib_devdata *dd)
 		int i;
 
 		dd->cspec->num_msix_entries = 0;
-		for (i = 0; i < n; i++)
-			free_irq(dd->cspec->msix_entries[i].vector,
-				 dd->cspec->msix_arg[i]);
+		for (i = 0; i < n; i++) {
+			irq_set_affinity_hint(
+			  dd->cspec->msix_entries[i].msix.vector, NULL);
+			free_cpumask_var(dd->cspec->msix_entries[i].mask);
+			free_irq(dd->cspec->msix_entries[i].msix.vector,
+			   dd->cspec->msix_entries[i].arg);
+		}
 		qib_nomsix(dd);
 	}
 	/* make sure no MSIx interrupts are left pending */
@@ -2597,7 +2600,6 @@ static void qib_setup_7322_cleanup(struct qib_devdata *dd)
 	kfree(dd->cspec->sendgrhchk);
 	kfree(dd->cspec->sendibchk);
 	kfree(dd->cspec->msix_entries);
-	kfree(dd->cspec->msix_arg);
 	for (i = 0; i < dd->num_pports; i++) {
 		unsigned long flags;
 		u32 mask = QSFP_GPIO_MOD_PRS_N |
@@ -3070,6 +3072,8 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 	int ret, i, msixnum;
 	u64 redirect[6];
 	u64 mask;
+	const struct cpumask *local_mask;
+	int firstcpu, secondcpu = 0, currrcvcpu = 0;
 
 	if (!dd->num_pports)
 		return;
@@ -3118,13 +3122,28 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 	memset(redirect, 0, sizeof redirect);
 	mask = ~0ULL;
 	msixnum = 0;
+	local_mask = cpumask_of_pcibus(dd->pcidev->bus);
+	firstcpu = cpumask_first(local_mask);
+	if (firstcpu >= nr_cpu_ids ||
+			cpumask_weight(local_mask) == num_online_cpus()) {
+		local_mask = topology_core_cpumask(0);
+		firstcpu = cpumask_first(local_mask);
+	}
+	if (firstcpu < nr_cpu_ids) {
+		secondcpu = cpumask_next(firstcpu, local_mask);
+		if (secondcpu >= nr_cpu_ids)
+			secondcpu = firstcpu;
+		currrcvcpu = secondcpu;
+	}
 	for (i = 0; msixnum < dd->cspec->num_msix_entries; i++) {
 		irq_handler_t handler;
-		const char *name;
 		void *arg;
 		u64 val;
 		int lsb, reg, sh;
 
+		dd->cspec->msix_entries[msixnum].
+			name[sizeof(dd->cspec->msix_entries[msixnum].name) - 1]
+			= '\0';
 		if (i < ARRAY_SIZE(irq_table)) {
 			if (irq_table[i].port) {
 				/* skip if for a non-configured port */
@@ -3135,7 +3154,11 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 				arg = dd;
 			lsb = irq_table[i].lsb;
 			handler = irq_table[i].handler;
-			name = irq_table[i].name;
+			snprintf(dd->cspec->msix_entries[msixnum].name,
+				sizeof(dd->cspec->msix_entries[msixnum].name)
+				 - 1,
+				QIB_DRV_NAME "%d%s", dd->unit,
+				irq_table[i].name);
 		} else {
 			unsigned ctxt;
 
@@ -3148,23 +3171,28 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 				continue;
 			lsb = QIB_I_RCVAVAIL_LSB + ctxt;
 			handler = qib_7322pintr;
-			name = QIB_DRV_NAME " (kctx)";
+			snprintf(dd->cspec->msix_entries[msixnum].name,
+				sizeof(dd->cspec->msix_entries[msixnum].name)
+				 - 1,
+				QIB_DRV_NAME "%d (kctx)", dd->unit);
 		}
-		ret = request_irq(dd->cspec->msix_entries[msixnum].vector,
-				  handler, 0, name, arg);
+		ret = request_irq(
+			dd->cspec->msix_entries[msixnum].msix.vector,
+			handler, 0, dd->cspec->msix_entries[msixnum].name,
+			arg);
 		if (ret) {
 			/*
 			 * Shouldn't happen since the enable said we could
 			 * have as many as we are trying to setup here.
 			 */
 			qib_dev_err(dd, "Couldn't setup MSIx "
-				    "interrupt (vec=%d, irq=%d): %d\n", msixnum,
-				    dd->cspec->msix_entries[msixnum].vector,
-				    ret);
+				"interrupt (vec=%d, irq=%d): %d\n", msixnum,
+				dd->cspec->msix_entries[msixnum].msix.vector,
+				ret);
 			qib_7322_nomsix(dd);
 			goto try_intx;
 		}
-		dd->cspec->msix_arg[msixnum] = arg;
+		dd->cspec->msix_entries[msixnum].arg = arg;
 		if (lsb >= 0) {
 			reg = lsb / IBA7322_REDIRECT_VEC_PER_REG;
 			sh = (lsb % IBA7322_REDIRECT_VEC_PER_REG) *
@@ -3174,6 +3202,25 @@ static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
 		}
 		val = qib_read_kreg64(dd, 2 * msixnum + 1 +
 			(QIB_7322_MsixTable_OFFS / sizeof(u64)));
+		if (firstcpu < nr_cpu_ids &&
+			zalloc_cpumask_var(
+				&dd->cspec->msix_entries[msixnum].mask,
+				GFP_KERNEL)) {
+			if (handler == qib_7322pintr) {
+				cpumask_set_cpu(currrcvcpu,
+					dd->cspec->msix_entries[msixnum].mask);
+				currrcvcpu = cpumask_next(currrcvcpu,
+					local_mask);
+				if (currrcvcpu >= nr_cpu_ids)
+					currrcvcpu = secondcpu;
+			} else {
+				cpumask_set_cpu(firstcpu,
+					dd->cspec->msix_entries[msixnum].mask);
+			}
+			irq_set_affinity_hint(
+				dd->cspec->msix_entries[msixnum].msix.vector,
+				dd->cspec->msix_entries[msixnum].mask);
+		}
 		msixnum++;
 	}
 	/* Initialize the vector mapping */
@@ -3365,7 +3412,7 @@ static int qib_do_7322_reset(struct qib_devdata *dd)
 	if (msix_entries) {
 		/* restore the MSIx vector address and data if saved above */
 		for (i = 0; i < msix_entries; i++) {
-			dd->cspec->msix_entries[i].entry = i;
+			dd->cspec->msix_entries[i].msix.entry = i;
 			if (!msix_vecsave || !msix_vecsave[2 * i])
 				continue;
 			qib_write_kreg(dd, 2 * i +
@@ -6865,15 +6912,13 @@ struct qib_devdata *qib_init_iba7322_funcs(struct pci_dev *pdev,
 
 	tabsize = actual_cnt;
 	dd->cspec->msix_entries = kmalloc(tabsize *
-			sizeof(struct msix_entry), GFP_KERNEL);
-	dd->cspec->msix_arg = kmalloc(tabsize *
-			sizeof(void *), GFP_KERNEL);
-	if (!dd->cspec->msix_entries || !dd->cspec->msix_arg) {
+			sizeof(struct qib_msix_entry), GFP_KERNEL);
+	if (!dd->cspec->msix_entries) {
 		qib_dev_err(dd, "No memory for MSIx table\n");
 		tabsize = 0;
 	}
 	for (i = 0; i < tabsize; i++)
-		dd->cspec->msix_entries[i].entry = i;
+		dd->cspec->msix_entries[i].msix.entry = i;
 
 	if (qib_pcie_params(dd, 8, &tabsize, dd->cspec->msix_entries))
 		qib_dev_err(dd, "Failed to setup PCIe or interrupts; "

drivers/infiniband/hw/qib/qib_pcie.c

@@ -194,11 +194,24 @@ void qib_pcie_ddcleanup(struct qib_devdata *dd)
 }
 
 static void qib_msix_setup(struct qib_devdata *dd, int pos, u32 *msixcnt,
-			   struct msix_entry *msix_entry)
+			   struct qib_msix_entry *qib_msix_entry)
 {
 	int ret;
 	u32 tabsize = 0;
 	u16 msix_flags;
+	struct msix_entry *msix_entry;
+	int i;
+
+	/* We can't pass qib_msix_entry array to qib_msix_setup
+	 * so use a dummy msix_entry array and copy the allocated
+	 * irq back to the qib_msix_entry array. */
+	msix_entry = kmalloc(*msixcnt * sizeof(*msix_entry), GFP_KERNEL);
+	if (!msix_entry) {
+		ret = -ENOMEM;
+		goto do_intx;
+	}
+	for (i = 0; i < *msixcnt; i++)
+		msix_entry[i] = qib_msix_entry[i].msix;
 
 	pci_read_config_word(dd->pcidev, pos + PCI_MSIX_FLAGS, &msix_flags);
 	tabsize = 1 + (msix_flags & PCI_MSIX_FLAGS_QSIZE);
@@ -209,11 +222,15 @@ static void qib_msix_setup(struct qib_devdata *dd, int pos, u32 *msixcnt,
 		tabsize = ret;
 		ret = pci_enable_msix(dd->pcidev, msix_entry, tabsize);
 	}
+do_intx:
 	if (ret) {
 		qib_dev_err(dd, "pci_enable_msix %d vectors failed: %d, "
 			    "falling back to INTx\n", tabsize, ret);
 		tabsize = 0;
 	}
+	for (i = 0; i < tabsize; i++)
+		qib_msix_entry[i].msix = msix_entry[i];
+	kfree(msix_entry);
 	*msixcnt = tabsize;
 
 	if (ret)
@@ -251,7 +268,7 @@ static int qib_msi_setup(struct qib_devdata *dd, int pos)
 }
 
 int qib_pcie_params(struct qib_devdata *dd, u32 minw, u32 *nent,
-		    struct msix_entry *entry)
+		    struct qib_msix_entry *entry)
 {
 	u16 linkstat, speed;
 	int pos = 0, pose, ret = 1;